Computer Vision

class towhee.hub.builtin.operators.computer_vision.image_load

load image from paths, file objects or memory blocks.

Returns

output image

Return type

ndarray

Examples:

>>> from towhee.hub import preclude
>>> from towhee.functional import DataCollection
>>> dc = (
...     DataCollection.range(5)
...         .tensor_random(shape=[100, 100, 3])
...         .image_dump()
... ).to_list()
>>> (
...     DataCollection(dc).image_load()
...         .map(lambda x: x.shape)
... ).to_list()
[(100, 100, 3), (100, 100, 3), (100, 100, 3), (100, 100, 3), (100, 100, 3)]
class towhee.hub.builtin.operators.computer_vision.image_dump(ext='.JPEG')[source]

dump image to a binary buffer.

class towhee.hub.builtin.operators.computer_vision.image_resize(dsize=None, fx=None, fy=None, interpolation=None)[source]

resize an image.

Parameters
  • dsize ((int, int), optional) – target image size. Defaults to None.

  • fx (float, optional) – scale factor for x axis. Defaults to None.

  • fy (float, optional) – scale factor for y axis. Defaults to None.

  • interpolation (str|int, optional) – interpolation method, see detailed document for cv2.resize. Defaults to None.

Returns

output image.

Return type

ndarray

Examples:

>>> from towhee.functional import DataCollection
>>> dc = (
...     DataCollection.range(5)
...         .tensor_random(shape=[100, 100, 3])
...         .image_resize(dsize=[10, 10], interpolation='nearest')
... )
>>> dc.select['shape']().as_raw().to_list()
[(10, 10, 3), (10, 10, 3), (10, 10, 3), (10, 10, 3), (10, 10, 3)]
class towhee.hub.builtin.operators.computer_vision.image_convert_color(code)[source]

convert image color space.

Parameters

code (str|int) – color space conversion string or code

Returns

output image.

Return type

ndarray

Examples:

>>> from towhee.functional import DataCollection
>>> import numpy as np
>>> (
...     DataCollection([np.ones([1,1, 3], dtype=np.uint8)])
...         .image_convert_color(code='rgb2gray')
...         .to_list()
... )
[array([[1]], dtype=uint8)]
class towhee.hub.builtin.operators.computer_vision.image_filter(ddepth, kernel)[source]

image filter.

class towhee.hub.builtin.operators.computer_vision.image_blur(ksize)[source]

image blur.

Parameters

ksize ([int]) – kernel size, for example: [3, 3].

Returns

output image

Return type

ndarray

>>> from towhee.functional import DataCollection
>>> import numpy as np
>>> (
...     DataCollection([np.ones([5,5], dtype=np.uint8)])
...         .image_blur(ksize=[3,3])
...         .to_list()
... )
[array([[1, 1, 1, 1, 1],
       [1, 1, 1, 1, 1],
       [1, 1, 1, 1, 1],
       [1, 1, 1, 1, 1],
       [1, 1, 1, 1, 1]], dtype=uint8)]